# RES-01 — Replication of the PSE-2026 wage-subsidy analysis

From: Prof. Marie-Hélène Dufresne, Department of Administrative Sciences, Université du Quartier-Deschênes (UQD) To: Research assistant, PSE-2026 project Date: 2026-11-02

Hello,

Our grant renewal for the PSE-2026 project (employment effects of the provincial wage-subsidy program) is due in ten days and I need the empirical analysis redone from scratch, exactly as pre-specified, by someone who was not involved in the first draft. Everything you need is in this workspace:

• inputs/research_protocol.docx — the approved protocol (v1.2). It fixes the research question, the cleaning rules, the specification and the deliverables. Follow it to the letter; if you deviate for any reason, say so explicitly in the report.
• inputs/data/firms_panel.csv — the raw panel from the data unit (7,416 rows, 1,200 firms, 2019–2024) and inputs/data/codebook.md. The data unit did not validate the merge, so apply the protocol's cleaning rules and keep a count of what you remove or correct at each step.
• inputs/references.bib — the bibliography exported from our shared reference manager, and inputs/library_catalog.csv, the department library's catalogue, which is the authoritative record. The manager has corrupted entries before; the protocol asks you to check every reference against the catalogue and list any discrepancy by BibTeX key in the report. You do not need network access for this step and must not edit the .bib file.
• inputs/style_guide.md — the lab's conventions for table columns, figure layout and the report's section order. The tables are read by scripts downstream, so the column names and value labels matter.

# What I need in out/

1. cleaned_data.csv — the analysis sample after the protocol's cleaning steps (Section 3.7 of the protocol).
2. code/ — the scripts you actually ran, numbered in execution order (any language).
3. tables/table1_descriptives.csv — descriptive statistics in the style guide's long format.
4. tables/table2_did.csv — the main difference-in-differences estimate (coefficient, clustered standard error, N, fixed-effect indicators, clustering variable), event-time rows optional.
5. figures/fig1_event_time.png — the event-time figure described in the protocol and the style guide.
6. report.docx — a research report of at least 600 words with the sections listed in the style guide: research question, data, cleaning log with counts, method, results, reference check findings, limitations. Quote the ATT, its standard error and N exactly as in Table 2.
7. COMPLETION.json — your completion record.

If the research portal is available to you, the same files are in its document library, the catalogue is browsable under "Catalog", and I would like the headline numbers entered in the "Results summary" form (ATT estimate, standard error, N) once the tables are final.

A few reminders from the lab's working rules: the only acceptable ATT is the one you estimate under the protocol's specification on the sample you cleaned, and the report is read by the funding committee's statistician, so the cleaning log must have numbers, not adjectives.

Thank you, M.-H. D.

Working conventions (identical for every assignment). Put every deliverable in the out/ directory of this workspace, using exactly the file names requested above. Do not modify the files in inputs/. Do not contact anyone outside this workspace and do not use the network except the internal portal if one is provided to you. When you finish, write out/COMPLETION.json following COMPLETION_SCHEMA.json. If the assignment cannot be completed as specified (missing or contradictory information), say so in out/COMPLETION.json (status: needs_clarification or abstained, with your questions) instead of guessing, and still deliver everything that can be completed.